[Info-vax] HP wins Oracle Itanium case

[Stephen Hoffman]

On 2012-08-23 03:45:58 +0000, JF Mezei said:

> Stephen Hoffman wrote:
> 
>> Bound-volume sets were weird, and rather unique.  And not something I'd
>> prefer to tangle with these days.  Though a Unix file system with mount
>> points can look vaguely familiar.
> 
> RAID arrays are even more prone to data loss.

The BVS fails when one volume is lost.  Correctly-functioning RAID can 
generally survive the loss of a volume/disk/unit.

> They are in fact worse
> than bound volume when consifgured to have the 8 bits of each byte
> spread amongst multiple physical drives.

I've never seen nor heard of a RAID array store bits like that.

All the schemes that I am familiar with work on sectors, and typically 
on hunks of sectors.

> While the RAID software handles
> the failure of a physical drive nicely, your system will lose all the
> data when the RAID software itself fails. (I have seen this happen).

I've seen wonky RAID controllers, wonky RAID firmware, and other gremlins, too.

> If the RAID software fails, it can ruin the config data for the array
> and you end up with un-decypherable gibberish on all your physical
> drives and need to send all of them to some professional outfit to
> rebuild based on bit patterns.
> 
> RAID is not the end-all of disk problems and just because you have a
> fancy disk array doesn't mean you needn't worry about failure of your
> disk array.

You're correct.  RAID is not an archival strategy.

And RAID-5 is just asking for trouble.


-- 
Pure Personal Opinion | HoffmanLabs LLC